Confluence metaheuristic: a novel initialization strategy for metaheuristic algorithms

• Published in: Evolving systems Vol. 15; no. 2; pp. 429 - 454
• Main Authors: Kaushik, Deepika, Nadeem, Mohammad
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.04.2024; Springer Nature B.V
• Subjects: Algorithms; Artificial Intelligence; Complex Systems; Complexity; Creeks & streams; Engineering; Genetic algorithms; Heuristic methods; Optimization; Original Paper; Particle swarm optimization; Problem solving; Sensitivity analysis; Statistical analysis; Streams; Metaheuristics; Confluence; Particle swarm optimization; Optimization; Genetic Algorithm
• ISSN: 1868-6478; 1868-6486
• DOI: 10.1007/s12530-023-09514-z

Metaheuristics represent an evolving problem-solving methodology that has proven to be effective in addressing optimization problems across various fields. Over the years, researchers have continuously strived to enhance the framework of metaheuristics in order to broaden their applicability, bolster their robustness, and ensure greater flexibility. In line with these objectives, we propose a novel framework called the Confluence Metaheuristic, which incorporates the concept of river confluence into the metaheuristic approach. The Confluence Metaheuristic operates by dividing the initial population into multiple streams, each subject to a different stochastic algorithm. These streams eventually converge at a single point and undergo further evolution through a specified number of iterations, yielding optimal results eventually. The proposed framework is of a general nature and has been implemented using two stochastic algorithms: particle swarm optimization (PSO) and genetic algorithm (GA). To validate the effectiveness of our proposed approach, we conducted extensive testing on ten benchmark functions. Furthermore, to substantiate the efficacy of our proposed idea, we subjected it to evaluation through three classical engineering problems. Statistical analysis was performed using Friedman and Wilcoxon tests, while sensitivity analysis was also conducted. The results obtained from this comprehensive evaluation were found to be highly satisfactory. Additionally, we compared the outcomes of our approach with the latest concept of ‘Multipopulation’. Moreover, our study contributes to improving various aspects of the metaheuristic framework, including the development of a more effective initialization strategy, a suitable parameter tuning approach, and an enhanced search technique.